Packfiles as efficient storage
Packfiles as efficient storage
Posted Feb 12, 2026 9:39 UTC (Thu) by MrWim (subscriber, #47432)Parent article: Evolving Git for the next decade
> Git's compression does not work well for binary files, and even small edits to such files means creating entirely new objects.
I’d have thought that pack files would be efficient at storing small changes to large files. As I understand it the format allows referring to ranges of content from existing objects (not in the same pack).
You still need to hash the whole file to name the new object, but you need to read the whole file either way to see if there are changes.
Can ‘git add’ write a pack file directly?